Surveyed Finance Leaders Spend 26% of Their Week Checking AI, Datarails Finds

AI has entered finance work, but human review remains the norm. In Datarails’ 2026 CFO Sentiments Survey, released October 6, 270 finance executives reported spending an average of 26% of their workweek checking or correcting AI output. Every respondent used AI, but only 5% trusted it to produce board-ready financial reports without review.

Global Surveyz Research conducted the online survey in July on behalf of Datarails, which sells AI finance software. Participants were recruited through a global research panel and worked at U.S. organizations with more than 1,000 employees and at least $100 million in annual revenue. The results reflect self-reported experiences from that group, not a measurement of every finance department.

The checking burden was widespread: 96% said they devoted at least 10% of their work time to verifying or correcting finance-specific AI outputs. For 8%, those checks consumed more than half their working time.

Wrong data, confident answers

The leading obstacle to trust was auditability—the ability to trace and check an output—which 75% cited for critical finance tasks. Accuracy and hallucinations, or invented answers, followed at 71%. Another 54% cited regulatory or compliance concerns.

Two recurring frustrations

• 65% identified confident answers based on the wrong data as their most common AI frustration.
• 56% said team members received materially different answers from a large language model despite using the same prompt and data.

The underlying data was not uniformly consolidated. Just 4% reported a single source of truth for finance and operational information. While 73% described their data as mostly centralized, 23% relied on disconnected systems and manual reconciliation—people matching records across systems.

Among teams whose leading operational challenges were manual reporting and data consolidation, 86% reported confident AI answers based on wrong data, versus 65% overall. That comparison links data-work difficulties with reported AI frustrations; it does not prove that disconnected systems caused the errors.

76%High implementation pressure

Respondents reporting high or very high pressure to fully implement AI.

7%Fully ready

Respondents saying their finance function was fully ready to implement AI across all workflows.

More licenses, not less oversight

The trust gap has not stopped purchasing plans. Of those surveyed, 53% planned to expand AI licenses over the following 12 months. Only 7% planned to cut or consolidate AI tools. Meanwhile, 32% said their organizations had exceeded AI budgets by at least 10% over the preceding year.

Those plans coexist with limited confidence in handing over sensitive work. Beyond board reports, just 4% trusted AI with month-end close, the process of finalizing a month’s financial records. The survey’s distinction is between using AI in finance and trusting it with these particular responsibilities.

Staffing responses point more toward reassignment than cuts within this sample. Sixty percent said they were moving staff to higher-value work as AI handled more everyday finance tasks. Just 3% said they were actively cutting finance-office staff because of AI. Those responses describe current staffing choices, not a forecast of future employment.